AI Memory Podcast Summaries
AI Memory on Yedapo: 5 summarized podcast and YouTube episodes. Each includes key takeaways, core concepts and notable quotes with timestamps.
Memory and Continual Learning: Engram's Dan Biderman and Jessy Lin
Sequoia Capital
Jun 24, 2026
Don Beerman and Jesse Lynn of Engram argue that relying on external retrieval (RAG) is a bottleneck for AI utility. They propose that models must move beyond static pre-training and external lookups to internalize company-specific context directly into their weights, enabling them to evolve alongside teams and perform complex tasks with significantly higher efficiency and lower token consumption.
Key insight: Engram’s founders suggest that internalizing context into model weights can reduce inference token consumption by up to 100x compared to traditional RAG-based approaches, as the model no longer needs to repeatedly process massive system prompts or search through external documents.
Build a Hermes Knowledge Base That Self-Improves
Jack Roberts
Jun 14, 2026
Current AI agents like Hermes suffer from 'amnesia,' limited only to conversational history. By integrating an Obsidian-based LLM Wiki, you can create a persistent, self-referential knowledge base that links your files, meetings, and research, effectively giving your AI a long-term memory that grows and improves independently.
Key insight: You can automate an AI agent to run daily background tasks that ingest meeting transcripts and external research directly into your Obsidian Wiki, ensuring your personal knowledge base scales exponentially without manual maintenance.
I Built The Best Claude Memory System (Beats Hermes)
Simon Scrapes
Jun 10, 2026
Agentic systems fail when they rely on single, limited memory frameworks. By synthesizing components from Memarch, Hermes, and GBrain, you can construct a hybrid system that handles automated storage, intelligent context injection, and semantic recall with verified citations.
Key insight: A memory system that confidentially admits it doesn't know an answer is fundamentally more valuable than one that hallucinates; true reliability requires citation-backed recall.
Every Claude Code Memory System Compared (So You Don't Have To)
Simon Scrapes
Apr 23, 2026
Effective AI agents require a structured memory system to prevent context rot and information loss. This episode breaks down six distinct levels of memory management, moving from native file-based storage to advanced semantic retrieval, helping you build a persistent 'business brain' that scales across your projects.
Key insight: If your Claude.md file exceeds 200 lines, you are likely suffering from 'context rot'; the solution is to use the file as an index that points to specialized, domain-specific external files rather than dumping all information into one place.
Claude Code Just Dropped Memory 2.0
Nate Herk | AI Automation
Mar 24, 2026
Anthropic’s unannounced AutoDream feature mimics human biology by periodically consolidating long-term memory files to eliminate cognitive bloat. By running background sub-agents to prune and merge context, Nate Herk argues that Claude ensures every new session feels sharp rather than cluttered. This shift transforms the AI from a simple chatbot into a platform with persistent, organized intelligence.
Key insight: The feature works like human sleep, using a background sub-agent to scan hundreds of past sessions and prune unnecessary tokens while preserving high-value project insights.